The Piedras Grandes unit

The Piedras Grandes unit consists of dominantly hornblende silicic andesite (60·1-64·0% SiO₂) with minor basaltic andesite bands and inclusions (up to 5 cm in diameter). Individual blocks show multiple, parallel, basaltic andesite bands 5-10 cm across. The andesite is poorly vesicular and porphyritic (42-47 vol. % phenocrysts; Table 1; Fig. 1). Phenocrysts are anhedral to euhedral. Plagioclase is the dominant phenocryst phase. Amphibole is the main ferromagnesian phase and ranges from fresh euhedral crystals to those with thin reaction rims of pyroxenes, plagioclase and Fe-Ti oxides (Fig. 1) to complete pseudomorphs. Minor phenocrysts of orthopyroxene, augite, Fe-Ti oxides and accessory apatite are present. Rare biotite crystals are commonly pseudomorphed by a fine-grained aggregate of pyroxenes, plagioclase and Fe-Ti oxides. Olivine crystals occur sparsely and are partially or wholly pseudomorphed by a vermicular orthopyroxene-magnetite symplectite. Anhydrite and scarce Fe-Cu sulphide and anhydrite inclusions occur in Fe-Ti oxide minerals. The groundmass consists of microphenocrysts of plagioclase, pyroxenes and Fe-Ti oxides set in a matrix of rhyolitic glass (Fig. 1).

Figure 1. Photomicrograph (crossed polars) of Piedras Grandes andesite, showing subhedral to euhedral, zoned plagioclase phenocrysts, both fresh and slightly reacted hornblende phenocrysts, and small, rounded orthopyroxene phenocrysts, in a glassy matrix containing plagioclase, orthopyroxene and hornblende microphenocrysts. Polars slightly uncrossed to show matrix texture. Field of view on long axis ~2 mm.

Table 1. Modal analyses of samples from the Piedras Grandes flow deposit, quoted on a vesicle-free basis

The mafic component is represented by a moderately phyric (12 vol. % crystals; Table 1) basaltic andesite band (LAS451a). The main phenocryst phases are plagioclase, olivine with Cr-spinel inclusions and augite with minor orthopyroxene, Fe-Ti oxides and accessory apatite. Occasional biotite and amphibole crystals are partially to completely pseudomorphed as in the host andesite. The matrix is a fine-grained groundmass of plagioclase, pyroxenes and Fe-Ti oxides.

Plagioclase phenocrysts in the andesite are euhedral, up to 2 mm in length (Fig. 1) and range from An₃₀ to An₇₁ with both normal and reverse zoned crystals (Fig. 2). In the basaltic andesite the compositional range of most phenocrysts is An_38-52, although one crystal with a calcic core (An₈₅) was found. Pyroxene phenocrysts in the andesite are subhedral to euhedral and up to 1 mm in length. Orthopyroxene phenocrysts in four samples of andesite have a restricted compositional range (mg-number = 0·65-0·69; Fig. 3) show no zoning of mg-number, apart from two crystals with a thin (10 µm) more magnesian overgrowth (mg-number = 0·72) and a magnesian inclusion (mg-number = 0·81) in an amphibole phenocryst. One sample of the andesite contains orthopyroxene with a wider compositional range (mg-number = 0·68-0·81) with both normal and reverse zonation. Orthopyroxene in a basaltic andesite band has a compositional range similar to that in the host andesite (mg-number = 0·67-0·70). Augite phenocrysts in the andesite (mg-number = 0·75-0·82), and a basaltic andesite band (mg-number = 0·78-0·82) are normally zoned. Olivine phenocrysts in the basaltic andesite are normally zoned, with core compositions of Fo_85-87 and rim compositions of Fo_79-80.

Figure 2. Plagioclase compositions in Piedras Grandes and Soncor rock types and Lascar Stage II biotite porphyries. `Xenolith' in a porphyry sample denotes coarse groundmass alkali feldspar between quartz xenocrysts.

Figure 3. Pyroxene compositions in Piedras Grandes andesite, Soncor ejecta (basaltic andesite, two-pyroxene pumice and hornblende pumice) and included rock types (hypabyssal lithic clasts, agglutinate and vitrophyre clasts).

Amphibole phenocrysts (up to 2 mm in length) fall mainly in the fields of tschermakite and tschermakitic hornblende (Leake, 1968). The amphibole phenocrysts show a wide range of mg-number from 2·9 to 3·5 in individual samples. Amphibole phenocrysts in the basaltic andesite are relatively Al poor and Si rich with a restricted range of mg-number (Fig. 4). Occasional amphibole crystals similar in composition to those in the andesite are interpreted as xenocrysts from the andesite. Biotite crystals in both the andesitic rocks and the basaltic andesite band are similar in composition and have a wide compositional range (Fig. 4).

Figure 4. Amphibole compositions from Piedras Grandes and Soncor (primitive and included) samples. Axis labels in formula units, calculated according to Leake, (1968). Element covariations are chosen that most effectively discriminate between different rock types.

The Soncor ejecta

The main juvenile components are two-pyroxene andesite and dacite pumice (61·0-67·6% SiO₂). These pumice clasts contain phenocrysts (28-44 vol. %, calculated vesicle free; Table 2) of plagioclase, orthopyroxene, augite and Fe-Ti oxides (see Fig. 6a, below) in a highly vesicular (24-68 vol. % vesicles) high-silica rhyolite glass matrix (76-77% SiO₂). Some samples contain minor amphibole phenocrysts and amphibole pseudomorphs consisting of plagioclase, pyroxene and Fe-Ti oxides. Completely fresh euhedral amphibole and coarse-grained pseudomorphs can occur in the same thin section. Sparse biotite is present in a few samples. Occasional rounded quartz, and olivine with reaction coronae of orthopyroxene-magnetite symplectite are present. Apatite is common as phenocrysts and as inclusions in phenocrysts. Rare zircon, anhydrite and pyrrhotite are present as inclusions in Fe-Ti oxides (not in the same individual crystals).

Table 2. Modal analyses of samples from the Soncor flow deposit, quoted on a vesicle-free basis

Hornblende andesite pumice clasts (58·0-63·8% SiO₂) from late ignimbrite flow units contain phenocrysts (32-39 vol. %, calculated vesicle free) of plagioclase, amphibole, orthopyroxene, clinopyroxene, Fe-Ti oxides and minor biotite (see Fig. 6b, below) in a low-silica rhyolitic glass matrix (71-74% SiO₂). One sample (LAS191) has a high-silica rhyolitic glass matrix (76-78% SiO₂). Crystal clots, involving various combinations of amphibole, plagioclase and orthopyroxene, are abundant. Also present are accessory apatite microphenocrysts, rare pyrrhotite inclusions in amphibole phenocrysts, and rare olivine crystals with orthopyroxene-magnetite reaction coronae. Some of these pumice clasts are heterogeneous because they vary locally from hornblende-rich, pyroxene-poor areas to areas with no hornblende and two pyroxenes similar in appearance to the main Soncor two-pyroxene pumice.

Basaltic andesite scoria (56·2% SiO₂) contains phenocrysts of olivine, augite, plagioclase, orthopyroxene and magnetite in an andesitic groundmass consisting of plagioclase, pyroxenes and magnetite, and rare ilmenite. Olivine crystals are rimmed with a fine overgrowth of augite, orthopyroxene, sub-calcic augite and pigeonite, and contain inclusions of Cr-spinel and andesitic glass. Augite crystals contain inclusions of orthopyroxene,pigeonite, magnetite and dacite glass.

Two varieties of glassy lithic clasts are interpreted as juvenile components. Glassy agglutinate clasts are interpreted as early welded pyroclastic facies of the Soncor eruption disrupted by later pyroclastic flows. They are dense, poorly vesicular, crystal-rich (40-50% crystals) hyalocrystalline silicic andesite. They contain phenocrysts of plagioclase, orthopyroxene, augite, amphibole and Fe-Ti oxides in a brown glassy andesitic matrix. They commonly contain lithic clasts of Stage I andesitic lavas, medium- to coarse-grained granitic rocks, rhyodacite porphyry and skarn xenoliths, as well as rounded and embayed quartz crystals. The matrix is heterogeneous with interbanded streaks of pale and dark brown glass. Dacite vitrophyres (65·7-67·7% SiO₂) are pale cream to pink, poorly vesicular welded rocks consisting of glassy fiamme and slightly flattened dacitic pumice. Phenocrysts and microphenocrysts of plagioclase, two-pyroxenes, amphibole and Fe-Ti oxides are contained in a cloudy, oxidized rhyolitic glass matrix.

Plagioclase phenocrysts in Soncor ejecta have large compositional ranges (An_32-84; Fig. 2) with complex zonation. The most calcic compositions (An_75-84) are found in phenocryst cores, which are commonly rounded and resorbed or sieve-textured. Oscillatory-zoned, more sodic overgrowths are common. A basaltic andesite scoria contains calcic cores (An_75-80) with more sodic (An_61-64) rims and microphenocrysts. Orthopyroxene and augite in all pumice types typically have very large compositional ranges (Fig. 3) with both normal and reverse zonation (mg-number opx = 0·65-0·84, mg-number cpx = 0·63-0·86). The entire compositional range can sometimes be observed in a single thin section. In contrast, the hornblende andesite pumices mostly lack augite and contain orthopyroxene with more restricted compositional ranges (mg-number = 0·66-0·75). Basaltic andesite scoria contains magnesian augite and orthopyroxene (mg-number opx = 0·76-0·79, mg-number cpx = 0·70-0·81) and normally zoned olivine (Fo_75-86). Amphibole phenocrysts in Soncor ejecta have a lower Fe³⁺/Fe²⁺ ratio, lower Al and higher Ti and Si than Piedras Grandes amphiboles, and more restricted range of mg-number (Fig. 4). One hornblende andesite pumice (LAS191) contains two compositional groups of amphiboles, one of which has lower Al and higher Si than the other group. Biotite phenocrysts in two-pyroxene pumice and agglutinates lie within the range of Piedras Grandes biotite compositions (Fig. 5). Biotite phenocrysts in one hornblende-rich pumice clast (LAS36-1) have relatively high Ti and low Al.

Figure 5. Biotite compositions from Piedras Grandes and Soncor (primitive and included) samples. Plots of Ti, Fe and Al against Mg (formula units calculated to 22 oxygens assuming all Fe as Fe²⁺).

Vent-derived lithic clasts

Vent-derived lithic clasts are divided into two main rock types: prismatic-jointed blocks and porphyries. These lithologies are interpreted to originate from the pre-Soncor Stage II volcanic complex and still hot subvolcanic intrusive bodies. They, together with the Piedras Grandes unit, provide information about the early stages of evolution of the magma system that eventually led to the zoned chamber, which was sampled by the Soncor eruption.

Prismatic-jointed blocks include both juvenile material and fragments of the pre-existing volcanic edifice. Two main varieties have been recognized.

Pale to dark green medium-grained basaltic andesites (56·1-56·8% SiO₂) are porphyritic or holocrystalline rocks with hypabyssal textures and containing phenocrysts of plagioclase, pyroxenes and Fe-Ti oxides (Fig. 6c). The matrix is an intersertal intergrowth of plagioclase, augite, orthopyroxene, Fe-Ti oxides, quartz and minor apatite. Euhedral plagioclase phenocrysts (up to 2 mm across) have large cores and oscillatory-zoned overgrowths. An andesitic prismatic-jointed block contains plagioclase with a restricted compositional range (An_40-50). Some crystals have sieve-textured cores and growth zones. Pyroxene phenocrysts (mg-number opx = 0·64-0·81, mg-number cpx = 0·74-0·86; Fig. 3) are up to 1 mm in length and euhedral. Orthopyroxene-magnetite intergrowths in crystal clots of orthopyroxene and plagioclase are interpreted as pseudomorphed olivine.

Figure 6. Photomicrographs. (a) Soncor two-pyroxene pumice (crossed polars), showing small plagioclase, orthopyroxene, clinopyroxene and magnetite phenocrysts in a vesicular glassy matrix. The central grey orthopyroxene has a clinopyroxene overgrowth. (b) Soncor hornblende andesite pumice (crossed polars), showing hornblende, orthopyroxene and plagioclase phenocrysts in a vesicular glassy matrix. (c) Basaltic andesite prismatic-jointed block (crossed polars) with flow-aligned plagioclase, orthopyroxene, corroded clinopyroxene and Fe-Ti oxides. In (a) and (b) the polars are slightly uncrossed to show matrix glass detail. (d) Type B porphyry (plane-polarized light), with corroded, oxidized biotite, sieve-textured plagioclase, and subhedral orthopyroxene phenocrysts in a medium-grained groundmass of plagioclase, pyroxenes and Fe-Ti oxides; Field of view on long axis of all photographs ~2 mm.

Pale two-pyroxene dacites (64·9-65·5% SiO₂) contain phenocrysts of plagioclase, pyroxenes, amphibole pseudomorphs, Fe-Ti oxides and occasional biotite in a matrix of colourless glass with microphenocrysts of plagioclase, pyroxenes and Fe-Ti oxides. Plagioclase phenocrysts (up to 4 mm in length) are subhedral to euhedral with large cores and oscillatory-zoned overgrowths. Occasional rounded, embayed and sieve-textured plagioclase crystals also occur. Pyroxene phenocrysts (mg-number opx = 0·64-0·81, mg-number cpx = 0·74-0·86; Fig. 3) are up to 2 mm in length and subhedral to euhedral. Biotite phenocrysts (up to 2 mm across) are rounded and embayed with a thin rim of Fe-Ti oxides. The mineral assemblage, mineral chemistries and textures are similar to Soncor pumice, and this rock type could represent an early dome of the Soncor eruption.

The porphyries are dense non-vesicular to poorly vesicular andesitic to dacitic rocks. They are divided into three petrographically distinct types.

Methods

One-pyroxene temperatures were calculated from individual orthopyroxene and clinopyroxene analyses using the `QUILF' program (Frost & Lindsley, 1992; Lindsley & Frost, 1992). This method projects the pyroxene compositions onto the pyroxene quadrilateral and calculates temperatures according to the method of Lindsley, (1983). Pressure dependence is slight (+2°C/kbar) and a pressure of 3 kbar was assumed. Aluminous pyroxenes (opx Al₂O₃ > 2% and cpx Al₂O₃ > 3 wt %) yielded variable and often very high temperatures, and were assumed to represent disequilibrium compositions produced under rapid-quench conditions (e.g. Schiffman & Lofgren, 1982; Ohnenstetter & Brown, 1992). Errors in the temperature calculations related to uncertainties in the probe analyses are estimated at ±10°C from repeated analyses of the same area of a crystal. The most important source of error is calibration of Si, which affects calculated Fe³⁺/Fe²⁺ ratios. Analyses of orthopyroxenes in experimental studies of the hornblende andesite of Montserrat (Barclay et al., 1998) produced calculated temperatures within 5°C of the run temperatures. However, Murphy et al., (2000) estimated an error of ±20°C in temperatures calculated for Soufriere Hills andesite, which is the product of 0·05-0·1 wt % errors in microprobe Ca analysis. We therefore estimate a maximum error of ±20°C in our calculations.

Temperature and oxygen fugacity were calculated from average titanomagnetite-ilmenite compositional pairs using the QUILF program (Andersen, 1993), which uses the model of Frost & Lindsley, (1992) and Lindsley & Frost, (1992). This method gave good agreement with glass inclusion melting temperatures in Lascar rocks (Matthews, 1994). At least 10 titanomagnetite and ilmenite crystals were analysed for each sample. Magnetite and ilmenite were checked for equilibrium using Mg-Mn partitioning (Bacon & Hirschmann, 1988), and only assemblages with equilibrium compositions, low compositional variability and absence of exsolution lamellae were used. This limited the calculations to rapidly cooled, usually dacitic rocks and very slowly cooled, re-equilibrated shallow intrusive rocks. For olivines with Cr-spinel inclusions in the basaltic andesite scoria, oxygen fugacities were calculated using the method of Ballhaus et al., (1990) and the temperature was calculated from augite one-pyroxene thermometry.

Results

The Piedras Grandes andesite displays a wide range of calculated temperatures (740°C to over 1060°C) despite the restricted range of pyroxene mg-number (Figs 7 and 8). Clinopyroxenes from a basaltic andesite band give temperatures of 1130-1220°C. Many individual orthopyroxenes have low-temperature cores and higher-temperature rims, with apparent temperature contrasts of up to 150°C. In contrast, the Soncor two-pyroxene pumices have a more restricted range of orthopyroxene temperatures (mostly 900-1000°C) over a wide range of mg-number (Fig. 8). Clinopyroxene temperatures fall in a similar range with a few high- and low-temperature outliers (Fig. 9). Orthopyroxene temperatures increase with increasing mg-number in Soncor pumice, albeit with considerable scatter (Fig. 8). Hornblende pumice samples overlap the Soncor field, but one sample (LAS191) gave lower and more varied temperatures (760-900°C) with a few outliers with lower temperature and lower mg-number. Core-to-rim zoning trends in Soncor pumice are both up- and down-temperature. The type A and B biotite porphyries gave a wide range of temperatures (740-950°C), up-temperature core-to-rim zoning trends and restricted mg-number, features similar to results from Piedras Grandes samples (Fig. 7).

Figure 7. Core-to-rim variations in temperature and mg-number for orthopyroxenes in Piedras Grandes andesite and basaltic andesite, and Soncor Type B porphyry lithic clasts, illustrating the similar compositional ranges and heating trends observed in orthopyroxenes in these rocks (marked by arrows connecting cores to rims).

Figure 8. Temperature-mg-number variations of orthopyroxene in Piedras Grandes and Soncor (primitive and included) ejecta. Data (upper diagram) and interpretation (lower diagram).

Figure 9. Temperature-mg-number variations of clinopyroxene in Piedras Grandes and Soncor (primitive and included) ejecta. Data (upper diagram) and interpretation (lower diagram).

The Piedras Grandes andesite, Soncor juvenile ejecta and vent-derived lithic clasts from the Soncor deposits contain equilibrium Fe-Ti oxide assemblages with low compositional variability. Three samples of Piedras Grandes andesite yielded temperatures of 905-925°C (Table 3). A basaltic andesite band recorded a slightly lower temperature (875°C). Soncor two-pyroxene pumices, hornblende pumices and a vitrophyre sample yielded temperatures in the range 860-940°C. Two andesitic scoria samples recorded a temperature of 900°C. Lower temperatures were calculated for an andesitic prismatic-jointed block (748°C) and two Type B biotite porphyry samples (740-766°C), and are interpreted as (probably sub-solidus) closure temperatures for magnetite-ilmenite equilibrium.

Table 3. Temperature and oxygen fugacity of Lascar ejecta from Fe-Ti oxide equilibria

Calculated oxygen fugacities of the Piedras Grandes and Soncor andesites and dacites range from FMQ (fayalite-magnetite-quartz) + 2 to + 2·9 (Table 3). High oxygen fugacity is supported by the ubiquitous presence of anhydrite in Piedras Grandes andesite, Soncor hornblende and two-pyroxene pumices, and vitrophyres (e.g. Carroll & Rutherford, 1987; Matthews et al., 1994a). Basaltic andesite samples range from FMQ + 1·1 to + 1·2, which is consistent with the interpretation of Matthews et al., (1994a) that oxygen fugacity increases with falling temperature in Lascar magmas.

Trace elements

Selected trace elements are plotted against SiO₂ in Fig. 11. Rb and Ba increase with increasing SiO₂, although Ba shows increasing scatter in more evolved compositions. V and Co decrease strongly with increasing SiO₂. Basaltic andesite magmas have highly variable Cr (48-230 ppm), Ni (17-112 ppm), V (130-190 ppm), Sr (440-606 ppm) and Zr (124-150 ppm), and in the case of V, Zr and Sr this variability extends to more silicic compositions. Cr and Ni are low for more silicic magmas (>60-61% SiO₂). Sr and Zr decrease markedly for SiO₂ contents higher than 60-62%. Lascar magmas have light rare earth element (LREE)-enriched patterns when normalized to chondrite, and Eu anomalies are weak or absent (Matthews et al., 1996).

Geochemical zonation and heterogeneity

The Soncor Plinian fallout deposit and ignimbrite show compositional heterogeneity and a weak but complex zoning pattern (Fig. 12). The main part of the Plinian deposit, as exposed on the south flank of Lascar, consists of white andesitic to dacitic two-pyroxene pumice with variable SiO₂ (62·5-66·5%). The uppermost 2 m of this deposit in the most proximal exposure also contains abundant banded pumice and dark andesitic pumice (61 wt % SiO₂) as well as white pumice, and exhibits a wide compositional range (61·0-67·3% SiO₂). The ignimbrite can be divided into two parts. The `main' flow units consist of lithic-rich lag breccias and pumice-richignimbrite. `Late' flow units are pumice- and scoria-rich distal flow units and the uppermost flow units on the SE flanks. The `main' flow units contain mostly andesitic to dacitic two-pyroxene pumice types (62·5-67·6% SiO<sub>2</sub>). The `late' flow units contain more silica-poor white pumice, compositionally banded pumice and andesitic scoria (59·2-65·0% SiO₂), hornblende andesite pumice types (58·0-63·8% SiO₂) and basaltic andesite scoria (56% SiO₂). A crude double zonation is indicated, with two cycles in which more mafic magma is erupted at the end of a cycle.

Figure 12. Whole-rock SiO₂ through the Soncor Plinian and later pyroclastic flow deposit showing complex compositional variation. The vertical axis represents relative stratigraphic position.

Glass compositions

Glass inclusions in olivine and clinopyroxene in a basaltic andesite (LA124) define a continuous compositional trend. Inclusions in olivines range from 56 to 62% SiO₂ and inclusions in clinopyroxenes range from 63 to 67% SiO₂. This trend shows variable TiO₂ and Na₂O, decreasing Al₂O₃, MgO and CaO, and increasing K₂O and P₂O₅. FeO* has a flat but variable trend to 63% SiO₂ then decreases sharply. The more silicic glass inclusions coexist with titanomagnetite.

Hornblende pumice samples contain low-silica rhyoliticglass, both as matrix and as inclusions in amphibole phenocrysts. Glass matrices and inclusions in phenocryst phases in two-pyroxene pumices are more evolved (76-77% SiO₂). Hornblende pumice sample LAS191 has matrix glass (and inclusions in minerals) similar in composition to that of the two-pyroxene pumices. There is no continuous compositional trend for all Soncor glasses. Glasses in hornblende and two-pyroxene pumices have separate compositional fields, both from each other and from the compositional trend of glass inclusions from basaltic andesite scoria. These low-silica rhyolitic glasses have relatively low TiO₂ and P₂O₅.

Chlorine and fluorine

In Soncor two-pyroxene and hornblende pumice samples apatites have approximately constant F/Cl ratio in most individual samples (Fig. 13a), but ratios differ between samples. Halogen contents increase towards apatite rims. Apatites from one sample of the Piedras Grandes andesite (SM93/10) show a trend of variable F at constant Cl content whereas another sample (SM93/20) shows no coherence (Fig. 13b). For most Soncor hornblende andesite pumices Cl is low in amphibole, with the exception of sample LAS191, which has relatively Cl-rich amphibole (Fig. 14a). This high-Cl trend is continued to higher mg-number and lower Cl by amphibole in a Type B porphyry. For both high- and low-Cl groups, Cl increases slightly with decreasing mg-number. Amphiboles in the Piedras Grandes andesite show low Cl in comparison with most Soncor ejecta (Fig. 14b). Individual samples of Piedras Grandes andesite contain hornblende phenocrysts covering a large range in mg-number, in contrast to Soncor samples, which show more restricted hornblende compositions in individual samples (Fig. 14). Biotite phenocrysts in Soncor two-pyroxene pumice have higher Cl contents than biotite in Piedras Grandes andesite and basaltic andesite (Fig. 15). Biotite in the Type B porphyries extends from the magnesian end of the Piedras Grandes range towards high-Mg, Cl-rich compositions. Glass inclusions in phenocrysts in Soncor hornblende and two-pyroxene pumices contain 0·10-0·16 wt % chlorine (Fig. 16a). Chlorine contents of matrix glasses are lower. Glass inclusions in olivine and clinopyroxene phenocrysts in the Soncor basaltic andesite scoria LA124 have more variable Cl (0·12-0·34 wt %).

Figure 13. F and Cl contents of (a) Soncor and (b) Piedras Grandes apatites. Formula units recalculated to 6(P + S + Si).

Figure 14. Cl contents of (a) Soncor and (b) Piedras Grandes amphiboles, plotted against Mg [formula units calculated according to Leake, (1968)]. The individual ranges of amphibole compositions in Soncor hornblende andesite pumices are clearly visible, as well as the relatively high Cl contents of amphiboles in Type B porphyries and the hornblende andesite LAS191. Piedras Grandes amphiboles have large compositional ranges in all samples.

Figure 15. Cl contents of biotites in Soncor and Piedras Grandes samples, plotted against Mg (formula units calculated to 22 oxygens, assuming all Fe as Fe²⁺). The rapid decrease in biotite Cl in the biotite porphyry samples, trending towards Piedras Grandes andesite biotite compositions, is clear. Soncor biotites have relatively high Cl contents.

Figure 16. (a) Cl contents of Soncor glass inclusions, showing a flat but scattered trend. The relatively large error bars on analyses of LA124 glass inclusions are due to analysis by microprobe using energy-dispersive spectrometry (EDS), as opposed to wavelength-dispersive spectrometry (WDS) for other samples. (b) S contents of Soncor glass inclusions, illustrating the very S-rich basaltic andesite glasses of Soncor olivine-rich scoria, and the rapid decrease in S with melt evolution.

Sulphur

Glass inclusions in phenocrysts in Soncor pumices (Fig. 16b) are relatively sulphur poor (<500 ppm). However, there is a negative correlation with glass SiO₂, indicating a melt composition control on sulphur solubility. Glass inclusions in olivine and clinopyroxene phenocrysts in the Soncor basaltic andesite show a strong decrease in S from 5000 ppm at 55% SiO₂ to <1000 ppm at 65% SiO₂.

Origin of mafic magmas

The variability of many major and trace elements in the basaltic andesite samples from the Piedras Grandes and Soncor eruptions implies that these magmas were modified before emplacement into the shallow magma system. Variations in Al₂O₃, Na₂O, Cr, Ni and Sr with little change in SiO₂ imply fractionation processes involving olivine, clinopyroxene and spinel. The abundance of these early phenocryst phases in the Soncor basaltic andesite sample LA124 is reflected in high Cr, Ni and V compared with other Lascar basaltic andesite samples. Some magmas became depleted in Cr and Ni by fractionation of olivine, clinopyroxene and Cr-spinel, whereas others (like LA124) entrained crystals of these phases and were thus enriched in these elements. The geochemical variations do not show any features indicative of plagioclase fractionation, which is consistent with evolution of basaltic andesite magmas by fractionation at high pressure. These features suggest that the basaltic andesite magmas were fractionated in the lower crust.

There is, however, poor correlation between Cr and Ni and other major and trace elements, indicating that fractionation and accumulation of olivine, clinopyroxene and spinel were not the only important processes. The LREE-enriched and relative heavy rare earth element (HREE)-depleted patterns of Lascar magmas implicate garnet (Kay et al., 1991). A number of models have been proposed for the `baseline' compositional and isotopic variability of Central Andean magmas, involving MASH (melting, assimilation, storage and homogenization) type processes (e.g. Hildreth & Moorbath, 1988; Rapp & Watson, 1995) in garnet-bearing lower crust. There is also evidence for modification of the mafic magma in the shallow magma chamber. For example, Mg-poor, low- to moderate-temperature (770-900°C) orthopyroxene phenocrysts and hornblende crystals identical to those in the andesite occur in the basaltic andesite band LAS451a of the Piedras Grandes unit. These phenocrysts are compositionally and thermally similar to the phenocrysts in the host andesite and are interpreted as crystals that have mixed into the basaltic andesite from the host magma.

Pre-Soncor magma evolution

The pre-Soncor Stage II volcano consisted of porphyritic hornblende andesites of the Piedras Grandes unit and porphyritic biotite dacite porphyries as sampled by the Soncor eruption. We interpret the petrological and geochemical data as the consequence of basaltic andesite magmas being emplaced into a high-temperature highly crystalline magma body that was remobilized by the supply of heat and to some extent mingled with the basaltic andesite.

Orthopyroxene crystals in the andesites of the Piedras Grandes unit and in the dacite porphyries have low-temperature (650-750°C) cores and show rising core-to-rim temperatures. We interpret the low-temperature orthopyroxene cores as restite crystals and the core-to-rim trends of rising crystallization temperature as evidence of reheating and partial melting in their origin. In the case of the Piedras Grandes andesite the presence of basaltic andesite bands and inclusions provides direct evidence that influx of hotter mafic magmas provided the heat input. Several features are consistent with hybridism between the andesite magma and the mafic magmas. Rounded, sieve-textured, anorthite-rich cores are observed in plagioclase crystals, commonly with more sodic, oscillatory zoned overgrowths. These cores have similar compositions to phenocrysts in the basaltic andesite. Rounded and embayed quartz grains with inclusions and fracture-fills of rhyolitic groundmass together with reacted olivine crystals in the andesite are illustrative of the disequilibrium hybrid character of the andesite. The amount of hybridism is, however, limited. The proportion of crystals that might originate from the basaltic andesite end member is low. The geochemical variations in the andesite do not fall on linear trends between more evolved dacitic rocks and the basaltic andesites as in other systems where hybridism can be demonstrated to be more significant (e.g. Clynne, 1999). The remobilization of highly crystalline low-temperature magma bodies by influx of mafic magmas with associated hybridism has been deduced in many other orogenic intermediate magmatic systems from similar kinds of evidence (e.g. Heiken & Eichelberger, 1980; Clynne, 1999; Murphy et al., 2000).

Fe-Ti oxides in Piedras Grandes rocks give consistent pre-eruption equilibration temperatures of 905-925°C, which are much higher than the temperatures given by the orthopyroxene cores but lower than temperatures indicated by some orthopyroxene rims and clinopyroxene crystals in the basaltic andesite inclusions. It is proposed that large temperature variations occurred in the Piedras Grandes andesite, but that thermal homogenization and equilibration occurred before eruption.

To discuss the formation of the Piedras Grandes andesite magmas and the dacite porphyries it is necessary to define the terms used. `Partial melting' is used here to indicate the formation of a partial melt by heating of a solid rock or partially molten rock that is too crystalline to behave as a magma. `Mobilization' is defined here as the formation of magma by partial melting. This magma consists not only of the partial melt, but also of entrained, unmelted, restite crystals. As recognized by Chappell & White, (1992), segregation of partial melt from restite crystals produces a magma of more evolved composition than the source rock. Complete mobilization produces a magma of the same composition as the source rock. The protolith could be crustal rocks unconnected genetically to the mafic magmas that provide the heat for mobilization. However, they could also be plutonic or cumulate rocks formed by previous batches of magma of the same character and origin. In this sense, distinguishing restite crystals from cumulate crystals is a matter of semantics. Situations in which previous magma batches have partly consolidated and are then remobilized and partiallymelted by later batches of magma in the same systemare likely to be very common.

We suggest that a solidified or largely solidified intrusion of andesite composition in the upper crust was activated at the beginning of Lascar Stage II by influx of basaltic andesite magma. This intrusion may represent the Lascar Stage I andesitic magma chamber, which cooled and solidified over a long period (thousands of years) of inactivity before the initiation of Stage II activity. The overall geochemical affinities of the Piedras Grandes andesite with the earlier Stage I andesites (Matthews et al., 1994b) are consistent with the ultimate origin of these magmas by fractional crystallization of basalt. Partial melting must have occurred at depths of at least 5 km to stabilize hornblende (Rutherford & Hill, 1993; Barclay et al., 1998). The orthopyroxene core temperatures (650-750°C) of the Piedras Grandes andesite and dacite porphyries indicate that this intrusion was close to its solidus temperature. However, it is possible that these temperatures are simply the final closure temperatures of the pyroxenes, preserved during cooling of the andesite intrusion, and that the actual temperature was even lower before remelting. The andesite rock was mobilized by mafic magmas with addition of heat and some mass. The dacite porphyries are interpreted as representing low-percentage partial melts with some entrained restite crystals. The Piedras Grandes andesite is interpreted to represent later complete mobilization of the intrusive protolith. Limited hybridism by the mixing of basaltic andesite with the mobilized andesite generated the geochemical trends in magmas more mafic than 61-62% SiO₂, and explains in part the inflections in geochemical trends at these compositions. This interpretation of some orogenic andesites to rhyolite sequences as the products of partial melting is increasingly supported by accumulated evidence, as exemplified by Mount St Helens (Blundy & Gardner, 1997), the andesite of Montserrat (Murphy et al., 1998, , 2000), Fish Canyon (Lipman et al., 1997) and granites of the Lachlan Fold Belt (e.g. Chappell & Stephens, 1988; Chappell & White, 1992; King et al., 1997).

The petrological features of the Piedras Grandes andesites can be interpreted in terms of fluid dynamical models of crustal melting (Huppert & Sparks, 1988). In mobilization of a solid or partially molten rock by mafic magma intrusions, melting occurs in the boundary layer region, but cooling and crystallization take place in the interior as its volume increases. The wall-rocks are variably heated both in space and time, and crystallization in the convecting interior occurs simultaneously with partial melting at the chamber margins. Parcels of magma with similar composition but with very different temperature and crystallization histories are progressively mixed in by convection to produce a fairly homogeneous magma, but with a highly heterogeneous crystal assemblage. The large spread of orthopyroxene temperatures and hornblende compositions in individual samples reflects this diverse thermal history. The more restricted Fe-Ti oxide temperatures are interpreted as the final temperature attained at the end of the remobilization process just before extrusion.

The heterogeneous Soncor chamber

The Soncor magma chamber contained a wide range of bulk magma compositions. The melt phase in Soncor two-pyroxene pumice clasts is always rhyolitic and so bulk magma compositional variation is largely a function of crystallinity. We interpret the bulk compositional Soncor ejecta in terms of fractional crystallization, but with significant disruption in the chamber as a result of convective stirring related to injection episodes of hydrous mafic magma at the base of the chamber.

The most mafic two-pyroxene pumice clasts have compositions comparable with the Piedras Grandes andesites. We interpret the Soncor two-pyroxene andesite and dacite magmas as products of fractionation of the observed phenocryst assemblage from an andesitic parent. The origin of the parent could be the melt composition formed by complete melting of the same source region that supplied the Piedras Grandes andesites, or more likely by formation of a hybrid between fractionated derivatives of the injected basaltic andesite magma and partial melts of the protolith. When mafic magmas are emplaced into a crustal magma system and exchange heat, magmas of intermediate composition are generated both by fractionation of the cooling mafic magma and by reheating of the pre-existing protolith (Huppert & Sparks, 1988). The exact contributions of partial melting and fractionation of cooling mafic magma are difficult to demonstrate conclusively, particularly if, as is thought to be the case at Lascar, the protolith and mafic magma are of the same magmatic lineage. We envisage the formation of higher-temperature andesitic magmas as a logical development of the process of magma chamber reheating by influx of mafic magma that initiated with the Piedras Grandes eruptions.

The phenocrysts and temperature estimates indicate that the Soncor magma chamber had a complex open-system history. The interpretation of its evolution is aided by reference to Zr and Sr variations (Fig. 17). In individual samples, orthopyroxene phenocrysts show a wide range of mg-numbers and calculated temperatures (mostly 900-1000°C). In some samples reverse and normally zoned orthopyroxenes suggest mingling of different magma parcels in the chamber. In most cases there is evidence for admixture of a small basaltic andesite component in these pumices, as evidenced by the occurrence of relict magnesian olivine crystals (replaced by magnetite-orthopyroxene symplectite), calcic plagioclase phenocrysts and some high-temperature (>1000°C) magnesian pyroxene crystals. These features are interpreted as evidence for influx of higher-temperature, more mafic magma, which stirred together parts of the fractionating chamber.

Figure 17. Interpretation of Sr and Zr patterns in Soncor magmas on the basis of a zoned magma chamber. Variable mafic magma compositions are interpreted to have mixed with an intermediate composition equivalent to the Piedras Grandes andesite. More evolved compositions are interpreted as products of fractional crystallization of this intermediate composition.

The origin of the hornblende pumice is now considered. These ejecta come from the last flow units of the ignimbrite and are assumed to be derived from the deeper levels of the chamber. Most of these samples have mafic andesite bulk compositions (58% SiO₂) and one possible interpretation is that they represent a melt composition from which the whole Soncor zoned system was derived. However, these samples contain abundant hornblende-plagioclase-orthopyroxene crystal clots as well as individual phenocrysts, suggesting that their mafic compositions can be partly attributed to admixture of cumulate materials. Observation of inhomogeneous pumice clasts containing both two-pyroxene and hornblende-rich domains indicates a hybrid relationship between the magmas.

Although the hornblende pumice might contain cumulate components, it cannot represent the cumulate equivalent of the main fractionated chamber, as such a cumulate would have to be relatively Zr rich rather than Zr poor (Fig. 17). A more probable explanation of the hornblende pumices is that they represent crystallization products of earlier episodes of hydrous basaltic andesite magma invading into the base of the Soncor chamber. This is supported by their low-silica rhyolite matrix glass compositions as opposed to the high-silica rhyolitic matrix glasses of the two-pyroxene pumices.

Hornblende was clearly stable in the lower parts of the Soncor magma chamber. The similar orthopyroxene temperature ranges (and magnetite-ilmenite temperatures) of the hornblende and two-pyroxene pumices suggest that hornblende stabilization was due to water pressure (e.g. Allen, 1978; Rutherford & Hill, 1993), and possibly also the influence of the somewhat more mafic compositions of these magmas. Recent experimental data on similar rock compositions imply that the chamber base would have to be at about 120-150 MPa (Gardner et al., 1995; Barclay et al., 1998), which would place the chamber in the upper crust at about 5-6 km depth. The occurrence of minor amounts of hornblende in some two-pyroxene pumices provides further support for a protracted open-system mixing history in the chamber. In the same samples hornblende ranges from coarse-grained plag-pyx-oxide pseudomorphs, to crystals with reaction rims, to perfectly fresh euhedral hornblende. The hornblendes have the same composition as in the hornblende pumices. We interpret the pseudomorphed hornblendes as the result of older periods of mixing of magma from deeper in the chamber, where hornblende was stable, into the upper parts, where it was unstable. The fresh hornblende may either be the consequence of the last episode of mixing associated with the eruption or it may be due to evolution of the temperatures and water pressure to a point where hornblende became marginally stable.

There is no direct evidence that the eruption of the Soncor magma chamber was triggered by a final magma injection event. Just before the eruption the Soncor magma had stabilized, and we take the Fe-Ti oxide temperatures as indicative of a thermal stratification in the chamber (860-940°C). The basaltic andesite scoria and pumice in the late flow units are interpreted as new hot magma emplaced at the base of the chamber. The petrology of the olivine basaltic andesite scoria implies slow cooling, which indicates that it had an extended residence time in the Soncor magma chamber. At least some of the heterogeneity in the ejecta can be associated with the convective disruption of the chamber and the conduit flow processes in the eruption. A complex magma chamber morphology is required for both hot basaltic andesite and hornblende andesite to exist at the base of the magma chamber. In one part, basaltic andesite was mixing with two-pyroxene andesite and dacite. In another part, hornblende andesite underlay the two-pyroxene magmas (Fig. 18).

Figure 18. Model for Piedras Grandes and Soncor magma chamber evolution. (a) Generation of biotite porphyries by remelting of andesitic protopluton, following injection of hot basaltic andesite magma. Basaltic hypabyssal rocks included in Soncor ejecta are products of slow cooling of this primitive magma at shallow depth. Green PJBs, green prismatically jointed rocks. (b) Total remobilization of protolith produces the Piedras Grandes hornblende andesite. Mafic inclusions are entrained by convection. (c) One possible configuration of the Soncor zoned magma chamber, with basaltic andesite, hornblende andesite and mixed magmas at the base and two-pyroxene dacite in the upper part. Fractional crystallization is dominant in the two-pyroxene dacite.

Interpretation of volatile behaviour

Chlorine and sulphur in the Piedras Grandes and Soncor magma chambers have complex histories. Low Cl contents of amphiboles and biotites in the Piedras Grandes andesites indicate low fHCl. Apatite trends of variable F at low Cl are attributed to open-system degassing of Cl during slow ascent of the Piedras Grandes andesite to the surface. Cl contents of amphibole and biotite in the Soncor two-pyroxene- and amphibole-bearing magmas indicate much higher fHCl than in the Piedras Grandes andesite. Poor homogenization occurred in the Soncor magma chamber, leading to differences in fHCl in different areas. Some glass inclusions from the basaltic andesite scoria are very Cl rich. We interpret these data to indicate that the basaltic andesite magma was the main source of Cl in the magma chamber. The approximately flat trend of Cl for glass inclusions with >65% SiO₂ in the Soncor pumices suggests probable melt saturation in a Cl-bearing volatile phase. We propose that Cl was steadily being degassed into a co-magmatic vapour phase during crystallization of this magma; this supports the interpretation that volatiles are transferred from mafic magmas into the overlying zoned chamber during replenishments. The trends of roughly constant F/Cl in apatite crystals in different Soncor pumice samples are consistent with an external buffer of halogen fugacity. We suggest that this buffer is the implied co-magmatic volatile phase.

Lascar is a sulphur-rich volcano, as indicated by the occurrence of anhydrite and by the large fluxes of SO₂ from the volcano (Andres et al., 1991; Matthews et al., 1994a, 1994b). The low solubility of sulphur in rhyolitic melts, as illustrated by analyses of sulphur in glass inclusions in Soncor ejecta, excludes the more evolved magmas as the major source for the sulphur. The olivine basaltic andesite (LA124) provides evidence that the basaltic magmas are the source, with melt inclusions in olivine containing up to 5000 ppm S. Crystallization of this magma caused exsolution of dissolved sulphur into a co-magmatic vapour phase, causing a rapid fall in melt S to <1000 ppm. The presence of anhydrite in the evolved Soncor magmas implies transfer of sulphur from invading mafic magmas in a free volatile phase (e.g. Andres et al., 1991; Matthews et al., 1994a) and requires both high fO₂ and high fS₂ (Carroll & Rutherford, 1987). Transfer of sulphur into the zoned chamber resulted in oxidation of S and dissolution into the melt as sulphate to stabilize anhydrite. The high fO₂ of Soncor and Piedras Grandes magmas resulted in oxidation of dissolved sulphur in the basaltic andesite to SO₂. Melt S in the Soncor magma chamber was dissolved dominantly as sulphate (Matthews et al., 1998). Buffering by a co-magmatic S-rich vapour was invoked by Matthews et al., (1994a) as the cause of the high fO₂.

The source of some of the Cl and all of the S is attributed to influxes of volatile-rich basaltic andesite magma, which exsolved a large proportion of its dissolved gas during quenching or slow crystallization after ponding at the base of the zoned Soncor magma chamber.

Water contents

The above discussion indicates that volatile-rich mafic magmas have been a major source of Cl and S during the evolution of the system. Water, however, is the major volatile species, and the transfer of water from invading hydrous basaltic magmas during partial melting or evolution of the Soncor zoned system is implied. Indeed, partial melting of largely solidified plutonic rocks cannot produce volatile-rich silicic magmas, as the volatile contents of the constituent minerals are too low. The requirement for substantial amounts of water to be transferred can be illustrated as follows.

The Piedras Grandes and Soncor magmas are interpreted to originate from remobilization of intrusive source rocks of andesite composition by invading mafic magmas. There is insufficient water in such rocks to stabilize the formation of hydrous phenocrysts in the resultant magmas. For example, let us consider a granodiorite porphyry sample as a possible protolith composition; magmatic water contents can be modelled during partial melting, assuming no loss of volatiles by degassing, or gain by addition from the injected and quenched mafic magma. The granodiorite porphyry has a low intrinsic water content, which is contained in amphibole. The amphibole content of this rock is 18 wt %, which yields a maximum total water content of 0·46 wt % (assuming 2·5 wt % H₂O in amphibole). If this composition is partially melted by 40%, a melt with 0·94 wt % H₂O could be formed. In both cases the water content is far too low for the stabilization of biotite and amphibole, and far too low in the case of the Soncor ejecta to account for a large-magnitude explosive eruption. Stabilization of amphibole in the Piedras Grandes and Soncor magmas requires water contents of about 4-5% (Gardner et al., 1995; Barclay et al., 1998). Low totals in glass inclusions from basaltic andesite and hornblende andesite ejecta (Table 6) are consistent with such water contents. Most of the Soncor ejecta are anhydrous, as a result of pressures being too low for stabilization of amphibole, but biotite is stable. Modelling of major Plinian eruptions requires water contents of at least 4% (Gardner et al., 1996; Melnik, 2000). These considerations imply transfer of volatiles as well as heat during influx of mafic magmas into evolved magma systems in the crust.